Building a reputation-based bootstrapping mechanism for newcomers in collaborative alert systems

Abstract

Today trust is a key factor in distributed and collaborative environments aimed to model participating entities behavior, and to foresee their further actions. Yet, prior to the first interaction of a newcomer in the system trust and reputation models face a great challenge: how to assign an accurate initial reputation to a newcomer? The answer needs to tackle two well-known problems: cold-start and reputation bootstrapping. Cold-start is a common issue to any system when newcomers boot for the first time, while reputation bootstrapping especially affects highly distributed scenarios, where mobile entities travel across domains and collaborate with a number of them. In this paper we focus on the two problems, which are addressed through a novel reputation bootstrapping mechanism for newcomers in a collaborative alert system aimed at detecting distributed threats. Experiments confirm the accuracy of our proposal as well as its robustness in the presence of ill-intentioned entities. © 2013 Elsevier Inc.